View
4
Download
0
Category
Preview:
Citation preview
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
1
IEEE FELLOW COMMITTEE
IEEE Fellow Strategic Planning Subcommittee (FSPS)
2016 REPORT
2016 FSPS voting members
1. Stefano Galli (FSPS Chair and IEEE Fellow Committee Vice-Chair)
2. Mariesa Crow
3. Serge Demidenko
4. Maria Sabrina Greco
5. Mark Karol
6. Dejan Milojicic
7. Amy Reibman (IEEE Fellow Committee Chair)
8. Tariq Samad
2016 FSPS non-voting members
1. Rosann Marosy (IEEE Fellow Activities Manager, staff)
2. Donna Dukes (IEEE Fellow Activities Administrator, staff)
From the IEEE Fellow Committee Operations Manual – Clause §2.1
The Fellow Strategic Planning Subcommittee is a standing committee of the IEEE Fellow
Committee, and is responsible for addressing ongoing and new activities on an annual basis,
including initiatives to improve the efficiency of and assure the fairness of the Fellow process,
encourage nomination of qualified individuals among underrepresented sectors of the IEEE
membership, and enhance the prestige of Fellow grade membership.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
2
THE 2016 FSPS REPORT
EXECUTIVE SUMMARY 4
ACKNOWLEDGEMENT 7
1. ITEM #1 – FINDING A SOLUTION TO THE PROBLEM OF THE EVER-INCREASING
NUMBER OF NOMINATIONS 8 1.1 The problem statement 8
1.2 The 2016 FSPS proposal on workload reduction 9
Previous approaches to reduce the Judges’ workload 10
Hiatus and Feedback for Nominees ranking in position #601 or below of the final IEEE ranking 11
Expedited Evaluation on Nominees ranking in Top and Bottom of S/TC rankings 12
1.3 The fundamental “selling points” of the proposal 13
Is a workload reduction procedure needed? 14
Do we risk increasing false positive/negatives? 14
Are we exercising insufficient oversight? 15
Is the proposal too radical? 15
Are there other approaches? 15
Is the proposed solution scalable? 16
What are the next steps? 16
1.4 The FSPS Motions for Item #1 (Workload Reduction) 17
2. ITEM #2 – REFORMING THE IEEE FC GOVERNANCE FRAMEWORK 19 2.1 The problem statement 19
2.2 The 2016 FSPS proposal on governance reform 20
Highlights of the Restructuring 20
Additional policy changes included in the revised Manual 22
2.2.2.1 The IEEE Fellow forms 23
2.2.2.2 Number of References 23
2.2.2.3 Appointment of FSPS members 23
2.2.2.4 S/TC-FEC term limits 24
2.2.2.5 S/TC-FEC Chair performs no evaluation 24
2.2.2.6 Governance requirements for S/TC-FECs 24
2.3 The FSPS Motions for Item #2 (Governance Reform) 25
3. ITEM #3 – IMPROVING COMMUNICATION WITH S/TCS 26 3.1 The FSPS Motions for Item #3 (Improving communication with S/TCs) 26
4. ITEM #4 – IMPROVING THE QUALITY OF NOMINATIONS 27 4.1 The FSPS Motions for Item #4 (Quality of Nominations) 27
5. ITEM #5 – INCREASING THE POOL OF APPLICANTS FROM UNDERREPRESENTED
DEMOGRAPHICS (GENDER, REGION, CATEGORY, ETC.) 28 5.1 Female Nominees 28
5.2 Industrial Nominees 29
5.3 The FSPS Motions for Item #5 (Underrepresented Demographics) 30
6. RECOMMENDATIONS TO NEXT YEAR’S FSPS 31
7. ANNEXES – ANALYSIS IN SUPPORT OF PROPOSALS 32 7.1 Analysis of efficacy and fairness of hiatus-based policies 33
Key Findings 33
Recommendation 34
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
3
Analysis of data 34
7.1.3.1 Method 1: Hiatus 37
7.1.3.2 38
7.1.3.3 Method 3: Hiatus of Y years based on being ranked in position #601 or below of the final IEEE
- 39
7.2 Analysis of the influence of the S/TC ranking on the final IEEE ranking 41
Key Findings 41
Recommendation 41
Analysis of data 41
7.2.3.1 IEEE rank changes 41
7.2.3.2 Correlation between S/TC and IEEE rank changes 43
7.3 Analysis of the elevation probability as a function of S/TC ranking 46
Key Findings 46
Recommendations 47
Analysis of data 47
7.3.3.1 Probability of elevation as a function of S/TC ranking 47
7.3.3.2 Nominees in the Top-T% of the S/TC rankings 51
7.4 Analysis on the elevation probability of specific categories of Nominees 53
Key findings with respect to the nominee’s EAT and NC 53
Key findings with respect to the nominee’s gender 55
Recommendations 56
Analysis of IEEE data 57
7.4.4.1 Nomination and elevation trends, by nomination category 57
7.4.4.2 Breakdown of employment affiliation types by nomination category 60
7.4.4.3 Elevation probabilities segmenting data by nomination category 61
7.4.4.4 Nomination and elevation trends, by employment affiliation type 66
7.4.4.5 Breakdown of nomination categories by employment affiliation type 69
7.4.4.6 Elevation probabilities segmenting data by employment affiliation type 71
7.4.4.7 Elevation probabilities segmenting data by both employment affiliation types and nomination
categories 77
7.4.4.8 Nomination and elevation trends, by nominee’s gender only 78
7.4.4.9 Elevation probabilities segmenting by nominee’s gender only 81
7.4.4.10 Breakdown of employment affiliation types of female nominees by nomination category 84
7.4.4.11 Breakdown of nomination categories of female nominees by employment affiliation type 86
7.4.4.12 87
7.4.4.13 89
7.4.4.14 Elevation probabilities segmenting data by both employment affiliation type and nomination
90
7.4.4.15 An interesting property unveiled in the analysis 91
Analysis of S/TC data 94
Considerations on the fairness of the Fellow process 96
8. REFERENCES 97
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
4
IEEE FELLOW COMMITTEE
IEEE Fellow Strategic Planning Subcommittee (FSPS)
2016 REPORT
Executive Summary
This report summarizes the activities of the 2016 IEEE Fellow Strategic Planning
Subcommittee (FSPS), a standing committee of the IEEE Fellow Committee (IEEE FC).
The members of the 2016 FSPS were appointed by the FSPS Chair at the beginning of the
calendar year. The FSPS transacted business via email and teleconference, discussed several
items of strategic importance, performed an extensive analysis of Fellow nomination and
elevation trends, and prepared several proposals to be presented at the October 2016 IEEE FC
meeting for approval.
Two high priority items have been identified by the FSPS and have been addressed thoroughly:
Item #1: Finding a solution to the problem of the ever-increasing number of nominations
Item #2: Reforming the IEEE FC governance framework
Additionally, the following important items have also been addressed:
Item #3: Improving communication w/Societies & Councils
Item #4: Improving the quality of nominations
Item #5: Increasing the pool of applicants from underrepresented demographics
Proposals for topics to be addressed by the 2017 FSPS are finally given in §6.
We summarize below the major findings and proposals for each of the five items listed above.
Item #1 – Finding a solution to the problem of the ever-increasing number of nominations
The number of nominations submitted every year has increased at an average rate of 3%/year
over the past 18 years, becoming a critical issue for the IEEE FC. Past FSPSs have addressed
this issue but most of the solutions that were proposed to reduce the workload on IEEE FC
Judges have been rejected by the IEEE FC.
The 2016 FSPS proposes a workload reduction procedure which has been carefully crafted by
performing an extensive data analysis and avoiding the pitfalls of past attempts. The workload
reduction proposal includes three interconnected elements:
1. Impose a hiatus for unsuccessful and low-ranked nominations
2. Provide feedback to low-ranked nominations
3. Introduce an “expedited review” for nominees ranked in the top and bottom of the
Society/Technical Council rankings.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
5
The supporting rationale for the above proposals is based on the extensive analysis on
nomination and elevation data reported in §7, where the analyses below are reported:
Analysis of efficacy and fairness of hiatus-based policies
Analysis of the influence of the S/TC ranking on the final IEEE ranking
Analysis of the elevation probability as a function of S/TC ranking
Analysis on the elevation probability of specific categories of Nominees
The above studies bring new insight into the problem of workload reduction, and also confirm
or dispel several “myths” about the Fellow evaluation process.
Item #2 – Reforming the IEEE FC governance framework
A variety of challenges make it difficult for the IEEE FC to implement changes in a timely and
efficient manner. To complicate things further, several details like the scoring procedure are
specified in the IEEE FC Manual, so that procedural changes adopted, for example, for coping
with the growing number of Nominees require IEEE Board of Directors approval.
With the primary goal of addressing the above issues, the 2016 FSPS has asked the advice of
the IEEE Governance Committee on establishing a new framework under which the IEEE FC
should operate in order to achieve better operational efficiency. The FSPS has also undertaken
the effort of mapping that advice to a revision of the 2015 IEEE FC Manual, while also tackling
ambiguities and missing provisions in the Manual currently in force.
The FSPS proposes to revise the way the IEEE FC handles its responsibilities and day-to-day
operations by clarifying in the Manual what actions or policies must be adopted with IEEE BoD
approval (hence specified in the Manual) or without it (hence specified in secondary governing
documents under the control of the IEEE FC). Furthermore, the revision also addresses what
responsibilities are under the entire IEEE FC and what responsibilities are delegated to its
(existing) standing subcommittee (Fellow Strategic Planning Subcommittee, or FSPS).
The above division of responsibilities has been codified in the following separate governing
documents (in order of precedence) each with its own approval authority:
1. The Operation Manual, requiring approval of IEEE FC and IEEE BoD
2. The document “Nomination and Evaluation Forms”, requiring approval of IEEE FC
3. Two new normative Handbooks requiring approval of IEEE FSPS and IEEE FC Chair:
a. “Additional Requirements, Responsibilities, and Guidelines for IEEE S/TC Fellow
Evaluating Committees”
b. “Fellow Evaluation Process and Scoring”
4. A new Recommendation Guide entitled “How to write an effective nomination,” which
requires approval of the IEEE FSPS and the Chair
5. A set of informative Web Help Guides (informative) to aid in the use of the IEEE web-
scoring program, maintained by the IEEE Fellow Activities Manager, and requiring
approval of the IEEE FC Chair.
IEEE FC approval will be requested for the above new/revised governance documents. These
documents are not included in this report and have been provided separately to the IEEE FC
(see file “Governance Files - Version FC.zip”).
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
6
Item #3 – Improving communication with S/TCs
If the workload reduction proposal in Item #1 is approved, the IEEE FC will deliberately exploit
the strength of the Society/Technical Council (S/TC) Fellow Evaluating Committees (FECs):
the capability of individuating the very strong and very weak Nominees. Even if the workload
reduction proposal in Item #1 is not approved, it is now well known that, even in the status quo,
the S/TCs exercise a good amount of influence on the choice of the very strong and very weak
Nominees – e.g., see the analyses in §7.2 and §7.3. Therefore, regardless of what happens to the
proposal in Item #1, the FSPS considered it very important to ensure that S/TC-FECs had:
4. Uniform and transparent practices
5. Enough rotation of FEC members, avoiding systematic bias
6. A good understanding of the tasks requested of them
The FSPS addressed the above requirements in a variety of ways, from adding to the IEEE FC
Manual governance requirements common to all S/TCs, to clarifying terms limits, to clarifying
in a new Handbook the evaluation tasks that S/TCs-FECs have to follow
Item #4 – Improving the quality of nominations
To address the objective of improving the quality of Fellow nominations, the FSPS crafted a
new Recommendation Guide entitled: “How to Write an Effective IEEE Fellow Nomination.”
The Guide provides examples and best practices from the perspective of IEEE Judges and S/TC
Evaluators. In addition to providing recommendations, this guide also includes a list of “things
to avoid” when completing the nomination. It is hoped that this Guide will help with the
submission of better written nominations, an indirect way of reducing the workload on Judges.
Item #5 – Increasing the pool of applicants from underrepresented demographics
The FSPS addressed the issue of increasing the nominations of underrepresented demographics,
and some of initiatives undertaken this year are described in the report. Furthermore, an in-
depth analysis on the nomination and elevation data of two important and underrepresented
categories (female and industry nominees) was performed analyzing trends, elevation
probabilities, and the potential presence of bias in the evaluation process.
Available data did not show evidence of unfairness in the Fellow process, neither between male
and female nominees nor between industry and academic nominees. As far as gender, it was
also found that the “elevation event” is somewhat1 independent of the nominee gender and that
elevations are being made proportionally to the a priori distribution of nominated men and
women. Furthermore, nominees in all employment affiliation types (industry, academia,
Government, other) have very similar elevation probabilities whereas Application
Engineer/Practitioner and especially Educator nominees experience a much lower elevation
probability when compared to Technical Leader and Research Engineer/Scientist nominees.
Industry and female nominees represent (1999-2016 temporal average) only the 26.4% and
5.7% of all nominees, respectively. In terms of trends, the trend of female nominations is
positive and, in the past five years (2013-2017), the percentage of female nominees has grown
1 Independence between “elevation event” and gender holds well for men but somewhat looser for women as
the wide yearly variability of female elevations does not allow, in certain cases, an estimate of the direct
conditional elevation probability as accurate as for the male case. For more details, see §7.4.2 and §7.4.4.9.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
7
to 7.4%. However, industry nominations are decreasing and the last 5-year average is now down
to 21.3%. Finally, it is pointed out that the share of female Fellows is today only 4.5% of all
Fellows, much lower than the 12.7% share of female IEEE members.
For the detailed analysis, additional results, and recommendations see §7.4.
Acknowledgement
The 2016 FSPS wishes to express its gratitude to the IEEE Fellow Activities Staff Rosann
Marosy and Donna Dukes without whom a lot of the work done this year would have been
impossible. Rosann and Donna were instrumental in gathering the Fellow data that enabled the
in-depth analysis presented in §7, giving the “historical” perspective of the IEEE FC, and
providing feedback on several of the 2016 FSPS initiatives.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
28
5. Item #5 – Increasing the pool of applicants from underrepresented
demographics (gender, region, category, etc.)
To increase the number of elevations of people from underrepresented demographics, the goal
was identified to increase the number of qualified candidates in our nomination pool. The
approach is to increase awareness of the Fellow process through the following activities. The
FSPS has looked in depth into how two specific categories do in the Fellow nomination and
elevation process: female and industrial nominees. For recommendations, see §7.4.3.
5.1 Female Nominees
A summary of activities and initiatives is given below:
Women Nominees:
o Advertisement in the “Women In Engineering (WIE)” newsletter, May/June
2016, see page 36 of the Newsletter
o Article in the WIE magazine, to appear in December 2016, interviewing the
IEEE FC Chair about the IEEE Fellow process.
o A Live Web chat, to be arranged.
o An analysis of nomination and elevation statistics for male and female nominees,
as well as a breakdown of male and female nominees in terms of nomination
categories and employment affiliation types, can be found in §7.4.4.8-§7.4.4.14.
The IEEE membership breakdown by gender and membership grades is shown in Table 2. Note
that percentages for women and men are calculated with respect to members with self-declared
gender only, while the percentage of Undeclared is calculated with respect to all members in
same grade.
Table 2 – Breakdown of IEEE membership by gender and grade
(F): Females; (M): Males; (U): Undisclosed Gender. Percentages of (F) and (M) are calculated ignoring (U).
Grade (F)% F (M)% M (U)% (U) Total
per grade
Total%
per grade
Student 30.3% 15,531 69.7% 35,716 28.5% 20,457 71,704 17.0%
Graduate Stud. 20.9% 6,493 79.1% 24,605 27.6% 11,832 42,930 10.2%
Member 8.7% 18,902 91.3% 198,103 14.1% 35,606 252,611 59.9%
Senior Member 6.4% 2,400 93.6% 35,018 3.0% 1,150 38,568 9.1%
Fellow 4.5% 339 95.5% 7,154 0.6% 49 7,542 1.8%
Associate 15.8% 996 84.2% 5,299 23.7% 1,958 8,253 2.0%
Honorary 4.3% 1 95.7% 22 28.1% 9 32 0.0%
Totals 12.7% 44,662 87.3% 305,917 16.9% 71,061 421,640 100.0%
An interest thing to note is that the percentage of female members decreases as female members
move up the membership grades. This decrease is monotonic, with women being 30.3% of
Student members and decreasing to 20.9% of Graduate Student members, 8.7% of Members,
6.4% of Senior Members, and finally to being only 4.5% of all Fellows. This does not happen
for male members whose share of membership grows monotonically with seniority of grade.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
29
Another interesting thing to note is that the 12.7% female membership in IEEE is actually
inflated by the high number of female students (graduate and non). If we eliminate all Student
Members except Graduate Student Members, then the share of women in IEEE goes down to
9.7%. If we eliminate all Student Members, then the share of women in IEEE goes down to
8.4%. This is much lower than the percentage of female engineers in the USA which a 2012
report of the U.S. Congress Joint Economic Committee (JEC) estimated to be at 14% – see the
report “STEM Education: Preparing for the Jobs of the Future”. This suggests that IEEE should
be doing a better job not only in attracting female engineers but also in retaining them as they
grow in their careers.
We also note that the percentage of female Senior members (6.4%) is on par with the share of
female Fellow nominees in recent years (7.3%, average 2012-2016). However, outreach efforts
for growing the number of qualified female nominees should be intensified anyway because the
number of female Fellows is only 4.5%.
The analysis of nomination and elevation statistics has uncovered interesting results which are
reported in §7.4.2. We report here the most important one: available data did not show evidence
of gender-bias in Fellow elevations, although it has also shown that the yearly data variability
for the female case does not allow, in certain cases, an estimate of the conditional elevation
probability for women as accurate as for the male case.
Specifically:
Averaging over 1999-2016, the direct conditional elevation probability of male (39%)
and female (39.3%) nominees is extremely close to the unconditional probability of
elevation (39%). This suggests that, on average, the “elevation event” for a nominee is
independent of gender. However, while the 95% Confidence Interval for the male
conditional elevation probability and the unconditional probability of elevation is only
2.3%, the one for the female conditional elevation probability is equal to 6.1% so that
event independence may hold looser for women. This means that the difference between
the male and female conditional elevation probabilities could be in the range ±8.4%.
Averaging over 1999-2016, the reverse conditional elevation probabilities (94.1% for
men and 5.9% for women) are extremely close to the a priori distribution of male
(94.3%) and female nominees (5.7%). The 95% Confidence Interval for all these
probabilities is around 1%, thus suggesting strong evidence for the fact that gender and
elevation are independent events and that, furthermore, elevations are being made
proportionally to the a priori distribution of male and female nominees.
5.2 Industrial Nominees
A summary of activities and initiatives is given below:
An exploration of the question whether Nominees from Industry or other employment
affiliation types (Academia, Government, Other) are disadvantaged or not was also
performed and key findings are summarized in §7.4.1. The statistical analysis indicates
that Nominees from industry actually have an average 8% higher probability of being
elevated when compared to Nominees in any other employment affiliation type
(Government, academia, and other). Available data did not show evidence of unfairness
in the Fellow process between nominees of any employment affiliation type. On the
other hand, an issue was found with Application Engineer/Practitioner and especially
Educator nominees who experience a much lower elevation probability than Technical
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
53
7.4 Analysis on the elevation probability of specific categories of Nominees
There is a perception that industry Nominees have a hard time becoming IEEE Fellows (harder
than academics) and that many more academics than industry members are getting elevated.
This analysis tries to shed some light on this matter and also tries to answer a more general
question: is there any category of members that has a better elevation probability than others?
In the analysis, membership has been segmented in order to categorize members in two ways:
By employment affiliation type (EAT): Industry (Ind.), Education/Academics (Acad.),
Government (Govt.), Other
By nomination category (NC): Application Engineer/Practitioner (AE/P), Educator
(EDU), Research Engineer/Scientist (RE/S), and Technical Leader (TL).
By gender – note that the information on the Fellow nominee gender is taken from the
IEEE membership database, not the Fellow Nomination form, and it is an optional field.
Thus, we here analyze the sets of self-declared men and self-declared women.
The analysis in §7.4.4 is based on IEEE data only. A similar analysis was performed on the data
of four Societies and is reported in §7.4.5. The Societies are: Computer (COMP), ComSoc
(COMM), Power & Energy (PES), and Signal Processing (SPS).
Available data for this study is from 1999 to 2016 for IEEE data segmented by EAT, 2007-2016
for IEEE data segmented by NC, and 2012-2016 for the considered Societies. Additional data
for the 2017 nominations was also used. Note that nominations and elevations of female
nominees are available since 1999; however, data on female nominees segmented by EAT and
NC is available only since 2012.
It is the first-time that such an extensive analysis is reported. Some limited analysis was done by
the 2009 FSPS, see Key Finding #3.
Key findings with respect to the nominee’s EAT and NC
The following key findings are noteworthy:
1. The number of Nominees employed in academia has more than doubled (+108%) in 1999-
2016, while the number of industry Nominees has decreased around 8% over the same
period. Today, academics account for around 71% of all nominations while Nominees
from industry account for around 21%, and Government plus Other for around 8%. This
means that academic Nominees constitute a much bigger talent pool than any other type of
Nominees and thus offers a bigger pool of talented candidates to choose from when
considering elevation. Furthermore, today 78% of Nominees are nominated in the RE/S
category, 11% as TLs, 7% as AE/P, and 4% as EDU.
2. When segmenting data based on NC, we find strong differences in elevation probability.
The analysis shows that members nominated in the RE/S category have an elevation
probability (averaged over time) that is +21% higher than the one of all other NCs
combined. Similarly, for TLs we have -3%, for AE/P -15%, and for EDU -36%. In
absolute numbers, the average elevation probabilities we found are: for RE/S it is 38.7%,
for TL it is 36.1%, for AE/P it is 31.8%, and for EDU it is 24.5%. Interestingly, the
categories with the lowest elevation probabilities are not constituted only of industry
members:
a. 95% of EDU Nominees are from academia (on the average).
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
54
b. 25% of AE/P Nominees are from the academia, Government, and Other while 75%
is from industry (on the average). However, note that in 2016, industry Nominees in
AE/P were only 60%.
3. The perception that industry members have lower elevation probability when compared to
academics is not substantiated by facts. The opposite is actually true: industry people have
a probability (averaged over time) of being elevated that is 8% higher than the elevation
probability of Nominees in all other EATs combined. On the other hand, academics have a
probability (averaged over time) of being elevated that is 5% lower when compared to all
other EATs combined. Overall, elevation probabilities of members segmented by EATs
are rather comparable: Industry: 41.3%, Other 39.6%, Government 39.5%, Academics
38.4% – this is very different from the NC case where we found that the average elevation
probabilities of EDU and AE/P Nominees were substantially lower than the ones of RE/S
and TL Nominees. The 2009 FSPS had looked at how industry Nominees were treated and
reached the following conclusion: “So, the issue it seems is not the way in which [industry]
nominations are treated, but rather the number of nominations received, and the challenge is
to increase the number of nominations in the category of Application Engineer/Practitioner”
[2009 FSPS]. The 2009 FSPS analysis looked only at two years of data (2008-2009) and,
at that time, did not investigate disparities related to the NCs which have been addressed
in this report (see Key Finding # 2).
4. Data excludes the presence of any “academia vs industry” large difference in terms of
elevation probability, and perception to the contrary may be fueled by the fact that
industry Nominees are in shorter supply than the academic one. A strong disparity in
elevation probabilities has actually been observed for the EDU and AE/P categories, and
especially for EDU which is strongly dominated by academics. This was also noted by the
2010 FSPS, looking at data from 2008-2010.
5. The success rates across categories differ strongly in the four S/TCs considered here
(Computer, ComSoc, Signal Processing, Power & Energy). The 2010 FSPS noted the
same in all S/TCs when considering data in 2008-2010.
6. Possible explanations for the lower elevation probabilities of EDU and AE/P:
a. Verifiable evidence and its impact are objectively more difficult to assess for AE/P
and EDU than for RE/S.
b. The talent pool of academics among the Nominees is much larger than the one of
any other EAT and the vast majority of Nominees are in the RE/S category, which is
easiest to evaluate in terms of verifiable evidence and impact. This contributes to the
erroneous perception that academics and researchers are favored. In reality, most
academics and researchers are nominated as RE/S which is perhaps the easiest
category to evaluate in terms of impact because the natural goal of research is to
produce public information with wide dissemination and its aim for impact has been
the object of scrutiny and quantitative assessment by government and private entities
for generations. Furthermore, academics are simply more numerous than any other
category and thus there are more good people to choose from.
7. It can be helpful to also consider externalities contributing to low percentages of
nominated industry members and low nominations rates in AE/P and TL may be. What
follows is not really a “Key Finding” but a set of well-educated opinions:
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
55
a. Many industry members are probably not engaged in activities for which Fellowship
is an appropriate recognition. For achievements in such activities, there are other
awards, medals, and various recognition instituted by the IEEE to recognize and
honor such types of accomplishments.
b. The Fellow grade is likely more important to academics. It is very well possible that
one of the reasons fewer industry candidates are nominated is that their company
gives them no or few incentives to seek a nomination, that elevation to Fellow is not
“rewarded” by their managers, and spending time and effort to go through the
process perhaps even frowned upon.
c. Corporate R&D has changed a lot since the golden years of a couple of generations
ago, and perhaps this has also to do with the fact that there many more incentives
and pressure for delivering short term results rather than investing in long term
R&D, e.g. the financial industry demands quarterly estimates, contract to executives
have short-medium duration and with incentives on short terms stock appreciation.
d. In support of the previous two points, we would like to cite a 2006 study done by
Robert Lucky and Jon Eisenberg for the USA National Academy of Engineering:
Renewing U.S. Telecommunications Research. The study showed a sharp change in
publication trends in telecom research between 1970 and 2005. In 1970, about 80%
of papers published in the IEEE Transactions on Communications were authored by
industry people, but in 2004 that number declined to a 15% (USA industry alone
went from 70% down to 7%). Similar trends hold also for conference papers (e.g.,
ICC and Globecom). The reduced contributions from industry have been partially
offset by an increase in the number of academic papers – both from U.S. and foreign
universities.
Key findings with respect to the nominee’s gender
The following key findings are noteworthy:
1. The 1999-2017 average number of female nominees is 5.8% of all nominations, although
in the five most recent years it has grown to an average of 7.3%. Although nominations in
the last decade have been in line with the fraction of female Senior Members in IEEE
(6.4% of all Senior members), earlier nominations were much fewer and in some years
could even be as low as zero.
2. Female fellows are 4.5% of all Fellows, which is well below the 12.7% of women in IEEE
(counting all membership grades).
3. Average (2012-2017) age of female nominees is 53.9 versus 56.3 for men. For female and
male nominees passed in 2012-2016 the average age is 53.5 and 56.3, respectively.
4. Available data did not show evidence of gender-bias in Fellow elevations, although it has
also shown that the yearly data variability for the female case does not allow, in certain
cases, an estimate of the conditional elevation probability for women as accurate as for the
male case. Specifically:
o Averaging over 1999-2016, the direct conditional elevation probability of male
(39%) and female (39.3%) nominees is extremely close to the unconditional
probability of elevation (39%). This suggests that, on average, the “elevation
event” for a nominee is independent of gender. However, while the 95%
Confidence Interval for the male conditional elevation probability and the
unconditional probability of elevation is only 2.3%, the one for the female
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
56
conditional elevation probability is equal to 6.1% so that even independence may
hold somewhat looser for women. The difference between the male and female
conditional elevation probabilities could be in the range ±8.4%.
o Averaging over 1999-2016, the reverse conditional elevation probabilities
(94.1% for men and 5.9% for women) are extremely close to the a priori
distribution of male (94.3%) and female nominees (5.7%). The 95% Confidence
Interval for all these probabilities is around 1%, thus suggesting strong evidence
for the fact that the gender and elevation are independent events and that
elevations are being made proportionally to the a priori distribution of male and
female nominees.
5. By far, female RE/S nominees are the largest group with an average of 50.3 nominees per
year. Female AE/P, EDU, and TL nominees only have a (2012-2017) average of 2.8, 3.3,
and 4.8 nominees per year.
6. By far, female academic nominees are the largest group with an average of 46.7 nominees
per year. Women in Other, Government, and industry have an (2012-2017) average of 0.8,
5.5, and 8.3 nominees per year.
7. Differently from the male case, AE/P, EDU, and RE/S are dominated by nominees in a
single EAT: industry, academics (100%), and again academics (82.1%), respectively.
Female TL nominees are equally distributed across academia, Government, and industry.
Recommendations
The following recommendations to the IEEE FC are made:
1. Although we have ascertained that the average elevation probability of industry members
is better that the one of academics, the number of industry Nominees has been declining
over the past 18 years. There can be many reasons for this, and the available data does not
allow pinpointing the root cause. However, the IEEE FC should intensify outreach
initiatives (e.g., via the Publicity Plan) to stimulate nominations from qualified industry
members. This will cause the small supply of AE/P and TL to increase, which will grow
the pool of good people to choose from, and finally raise their elevation probabilities.
2. One of the identified possible causes for the lower elevation probability of EDU and AE/P
Nominees compared to RE/S ones is the objective difficulty of (a) making the case in
terms of verifiable evidence and its impact and (b) assessing the evidence presented. The
IEEE FC should undertake the following actions to address these issues:
a. Increase the quality of nominations, Reference, and endorsements by providing clear
guidelines and recommendations to all the participants in the Fellow process. This
will help improving all nominations, and thus also the AE/P and EDU ones.
b. Improve the orientations courses for Society Evaluators and IEEE Judges, and
specifically address the issue of how to weigh the presented evidence and how to
assess it in a way that is meaningful to the specific NC of the Nominee.
3. Outreach initiatives to stimulate the number of qualified female Nominees should continue
and be intensified in order to increase the number of female nominees and also stabilize
the trend of elevations which today continues to exhibit high variability between years.
Additional efforts should be directed at increasing the diversity of nominees in terms of
EAT as today the vast majority of female nominees is from academia.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
57
4. The Fellow Committee should have diverse membership in terms of gender and other
underrepresented demographics. Fellow Committee members should also actively solicit
nominations for qualified people that are underrepresented in the Fellows category, and
especially women who today are only 4.5% of all Fellows – largely below the 12.7% of
female members in the IEEE.
Analysis of IEEE data
The analysis is organized in several sections, each of which focuses on different aspects and
subsets of the data:
1. Data segmented by NCs
a. Trends of elevations and nominations over time
b. A breakdown of EATss in terms of NCs
c. Elevation probabilities segmenting data by NC
2. Data segmented by EATs
a. Trends of elevations and nominations over time
b. A breakdown of NCs in terms of EATs
c. Elevation probabilities segmenting data by EAT
3. Data segmented by both EATs and NCs
4. Data segmented by Nominee’s gender
a. Trends of elevations and nominations over time, for female Nominees only
b. Elevation probabilities for female Nominees only
c. A breakdown of EATs in terms of NCs, for female Nominees only
d. A breakdown of NCs in terms of EATs, for female Nominees only
e. Elevation probabilities segmenting data by NC, for female Nominees only
f. Elevation probabilities segmenting data by EAT, for female Nominees only
g. Elevation probabilities segmenting data by EAT and NC , for female Nominees only
5. An interesting property of the data
When relevant, the 95% Confidence Interval (95%-CI) in the estimate of the temporal mean of
the conditional elevation probability will also be reported.
7.4.4.1 Nomination and elevation trends, by nomination category
Let us now look at the data by segmenting it in terms of NCs. This is shown in Figure 4, where
it is easy to confirm that the RE/S Nominees continue to grow and they are by far the single
dominating NC of all both in terms of nominations and elevations. The percentage of
nominations of each NC will be shown later, see Table 27.
In Figure 5, we can appreciate better the trend of AE/P, EDU, and TL. Compared to RE/S, the
number of Nominees in these three NCs has remained pretty constant: EDU nominations
decreased around 10% and AE/P nominations increased around 10%. On the other hand, while
elevations of AE/P and TL Nominees have grown around 10%, the elevations of EDU
Nominees have decreased as much as 50% over the past decade.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
78
Table 42 – Conditional probabilities of elevation { | }, min|mean|max for 2007-2016. The last
column on the right contains { | } (see Table 63, left side) while the bottom row contains { | }
(see Table 63, right side), both averaged over 2007-2016.
AE/P EDU TL RE/S
Ind 20.0%|32.5%|42.6% 0.0%|28.6%|100% 31.0%|39.0%|46.3% 28.9%|44.2%|58.2% 39.7%
Acad 0.0%|31.9%|66.7% 17.8%|25.2%|42.4% 20.0%|31.6%|64.7% 32.7%37.5%|43.3% 36.4%
Govt 0.0%|22.3%|50.0% 0.0% 7.1%|32.7%|53.8% 30.3%|41.8%|56.5% 37.2%
Other 0.0%|49.2%|50.0% 0.0% 0.0%|33.3%|75.9% 11.1%|39.9%|66.7% 37.9%
31.8% 24.5% 36.1% 38.7%
The table shows several interesting things:
None of Government and Other members nominated in EDU have ever been elevated
Some subsets of Nominees have conditional elevation probabilities that have very wide
ranges, such as Industry members nominated as EDU, Academics nominated as AE/P,
Government and Others nominated as AE/P, etc.
Some subsets of Nominees have conditional elevation probabilities that have narrower
ranges, such as Industry members nominated as AE/P or TL, Academics nominated as
RE/S, etc.
Industry Nominees have the highest average conditional elevation probability across
every EAT and every NC.
7.4.4.8 Nomination and elevation trends, by nominee’s gender only
The Nominee’s gender is not included in the Nomination form and it is taken from the IEEE
membership database, where it is an optional field. IEEE has data on male/female nominees as
well as nominees of undisclosed gender since 2012, but we have no information on nominees of
undisclosed gender between 1999 and 2011. In our analysis, we will approximate the set of
male nominees with the set of all nominees minus female nominees in years 1999-2011, thus
assuming that prior to 2012 there were no nominees of undisclosed gender. As shown in Table
2, the percentage of members of undisclosed gender for the two membership grades of interest
to the Fellow process (Senior and Fellow) is very low. This makes plausible approximating the
number of nominees of undisclosed gender to zero when data is not available.
Figure 11 shows the nomination and elevation trends for all Nominees and also for male
nominees. We can see that the trend of male nominees closely follows the trend of all nominees,
if not for the growing (yet small) gap due to the increasing number of female nominees over
time.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
79
Figure 11: Nomination and elevations trends of male nominees vis-à-vis all nominees.
The trend for female nominees is shown in Figure 12. We note a sustained growth in
nominations from 21 in 1999 to 80 in 2017 (+281%) and a corresponding growth in elevations,
from 13 in 1999 to 23 in 2016 (+77%), which however is around four times slower than the
growth in nominations.
For the case of female elevations, the IEEE has data that goes back to the mid 60s. This data is
plotted in Figure 13 which shows that the number of elevations of female nominees did not
really start to grow until the early 90s. Furthermore, it is interesting to note that the number of
female elevations has never been smooth and continues to exhibit a high variability over time
also in recent years. This variability of elevations coupled with the small sample size will cause
wide variations in the conditional elevation probability and, as a consequence, a widening in the
95% Confidence Interval (95%-CI) with respect to the male case.
0
100
200
300
400
500
600
700
800
900
1000Male Nominees and All Nominees
All Nominations
Male Nominations
All Elevations
Male Elevations
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
80
Figure 12: Nomination and elevations trends of female nominees.
Figure 13: Historical plot of the number of elevations of female nominees.
0
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85 Female Nominees
Nominations
Elevations
0
5
10
15
20
25
30
1966 1976 1986 1996 2006 2016
Female Elevations
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
81
7.4.4.9 Elevation probabilities segmenting by nominee’s gender only
The available data is shown in Table 43. All the results discussed in this section can be obtained
starting from this table.
Table 43 – Nominations and elevations by gender.
Unsuccessful Nominations Successful Nominations
Class of Males Females Total F Males Females Total P Total F&P
1999 319 8 327 226 13 239 566
2000 279 4 283 246 2 248 531
2001 257 12 269 251 5 256 525
2002 287 15 302 246 13 259 561
2003 351 18 369 246 14 260 629
2004 385 30 415 254 6 260 675
2005 481 29 510 251 17 268 778
2006 477 37 514 264 7 271 785
2007 470 30 500 247 18 265 765
2008 459 20 479 267 27 294 773
2009 432 27 459 279 19 298 757
2010 457 35 492 283 22 305 797
2011 471 23 494 290 29 319 813
2012 433 29 462 301 23 324 786
2013 490 37 527 272 19 291 818
2014 507 42 549 268 19 287 836
2015 530 33 563 265 26 291 854
2016 490 37 527 265 23 288 815
Mean 420.8 25.9 446.7 262.3 16.8 279.1 725.8
Following the same methodology of §7.4.4.3 and §7.4.4.6, we obtain the marginal probabilities
of Table 44 and the direct and reverse conditional probabilities of Table 45.
Table 44 – Marginal probabilities of male and female nominees, and successful P(P) and unsuccessful P(F)
nominations. The last two-rows show column-wise (temporal) averages and the 95%-CI of the average,
respectively.
Class of P(Male) P(Female) P(F) P(P)
1999 96.3% 3.7% 57.8% 42.2%
2000 98.9% 1.1% 53.3% 46.7%
2001 96.8% 3.2% 51.2% 48.8%
2002 95.0% 5.0% 53.8% 46.2%
2003 94.9% 5.1% 58.7% 41.3%
2004 94.7% 5.3% 61.5% 38.5%
2005 94.1% 5.9% 65.6% 34.4%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
82
Class of P(Male) P(Female) P(F) P(P)
2006 94.4% 5.6% 65.5% 34.5%
2007 93.7% 6.3% 65.4% 34.6%
2008 93.9% 6.1% 62.0% 38.0%
2009 93.9% 6.1% 60.6% 39.4%
2010 92.8% 7.2% 61.7% 38.3%
2011 93.6% 6.4% 60.8% 39.2%
2012 93.4% 6.6% 58.8% 41.2%
2013 93.2% 6.8% 64.4% 35.6%
2014 92.7% 7.3% 65.7% 34.3%
2015 93.1% 6.9% 65.9% 34.1%
2016 92.6% 7.4% 64.7% 35.3%
Mean 94.3% 5.7% 61.0% 39.0%
95%-CI ±0.8% ±0.8% ±2.3% ±2.3%
We point out that the P(F) and P(P) in the above table are very close but not the same as the one
shown in Table 37. The reason they are not the same is that the nominees of undeclared gender
have been excluded in the above table while all nominees (included the ones with an
undisclosed gender) were accounted for in Table 37.
Table 45 – Direct and reverse conditional elevation probabilities, when partitioning by NC (women only).
The last two-rows show column-wise (temporal) averages and the 95%-CI of the average, respectively.
Class of P(P|Male) P(P|Female) P(Male|P) P(Female|P)
1999 41.5% 61.9% 94.6% 5.4%
2000 46.9% 33.3% 99.2% 0.8%
2001 49.4% 29.4% 98.0% 2.0%
2002 46.2% 46.4% 95.0% 5.0%
2003 41.2% 43.8% 94.6% 5.4%
2004 39.7% 16.7% 97.7% 2.3%
2005 34.3% 37.0% 93.7% 6.3%
2006 35.6% 15.9% 97.4% 2.6%
2007 34.4% 37.5% 93.2% 6.8%
2008 36.8% 57.4% 90.8% 9.2%
2009 39.2% 41.3% 93.6% 6.4%
2010 38.2% 38.6% 92.8% 7.2%
2011 38.1% 55.8% 90.9% 9.1%
2012 41.0% 44.2% 92.9% 7.1%
2013 35.7% 33.9% 93.5% 6.5%
2014 34.6% 31.1% 93.4% 6.6%
2015 33.3% 44.1% 91.1% 8.9%
2016 35.1% 38.3% 92.0% 8.0%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
83
Class of P(P|Male) P(P|Female) P(Male|P) P(Female|P)
Mean 39.0% 39.3% 94.1% 5.9%
95-CI% ±2.3% ±6.1% ±1.2% ±1.2%
On the average, the conditional elevation probability of male and female nominees is almost the
same. Furthermore, the probability of finding a male or a female among the elevated nominees
is very close to the a priori unconditional probabilities of having a male or female nominee, thus
suggesting that elevations are being made proportionally to the a priori distribution of men and
women among the nominated population.
The average (temporal) elevation probability for male nominees is reasonably accurate and has
been calculated to be equal to 39% with a 95% Confidence Interval (CI) of ±2.3%. The average
(temporal) elevation probability for female nominees has been calculated to be equal to 39.3%
but is affected by a much larger CI equal to 6.1%. This higher CI is due to the high variability
of the elevation probability over time.
The metric “edge” as defined in §7.4.4.3 boils down, in this case, to the ratio of the
probabilities of the two left-most columns of Table 45 minus 1. Edge is shown in Table 46while
a plot of edge is also shown in Figure 14.
Table 46 – The edge for male and female nominees. The last two-rows show column-wise (temporal)
averages and the 95%-CI of the average, respectively.
EDGE Male Female
1999 -0.33 0.49
2000 0.41 -0.29
2001 0.68 -0.40
2002 -0.01 0.01
2003 -0.06 0.06
2004 1.38 -0.58
2005 -0.07 0.08
2006 1.24 -0.55
2007 -0.08 0.09
2008 -0.36 0.56
2009 -0.05 0.05
2010 -0.01 0.01
2011 -0.32 0.46
2012 -0.07 0.08
2013 0.05 -0.05
2014 0.11 -0.10
2015 -0.24 0.32
2016 -0.08 0.09
Mean 0.12 0.02
95%-CI ±0.25 ±0.16
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
84
Figure 14: Plot of “edge” metric for male and female nominees.
The analysis of edge confirms that, despite wide yearly variations exemplified in the oscillatory
behavior, the two categories of male and female nominees have similar average conditional
elevation probability, with an historical higher edge for men. However, in recent years, the edge
of women has increased and became larger than the one of men which actually became
negative. For example, averaging over the past 10 years yields =-0.11 for men and =0.15 for
women; averaging over the past 5 years yields =-0.05 for men and =0.07 for women. It is
also noteworthy that the yearly variations of edge seem to be decreasing in recent times.
7.4.4.10 Breakdown of employment affiliation types of female nominees by
nomination category
The breakdown is shown in the next four tables, from Table 47 to Table 50. Given the small
number of female nominees in some EAT/NC as well as the availability of only 5 samples
(years), the averages reported for non-academics must be taken with a grain of salt.
Table 47 – Composition of Industry Nominees by NC (female only).
Ind AE/P EDU TL RE/S
2012 12.5% 0.0% 12.5% 75.0%
2013 14.3% 0.0% 28.6% 57.1%
2014 50.0% 0.0% 0.0% 50.0%
2015 33.3% 0.0% 22.2% 44.4%
2016 28.6% 0.0% 14.3% 57.1%
-1.00
-0.50
0.00
0.50
1.00
1.50EDGE (Male and Female Nominees)
Male
Female
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
85
Ind AE/P EDU TL RE/S
2017 23.1% 0.0% 23.1% 53.8%
Mean 27.0% 0.0% 16.8% 56.3%
Table 48 – Composition of Academic Nominees by NC (female only).
Acad AE/P EDU TL RE/S
2012 0.0% 5.1% 5.1% 89.7%
2013 2.3% 11.4% 2.3% 84.1%
2014 0.0% 6.0% 2.0% 92.0%
2015 0.0% 11.4% 0.0% 88.6%
2016 2.3% 4.7% 4.7% 88.4%
2017 1.7% 5.0% 5.0% 88.3%
Mean 1.0% 7.3% 3.2% 88.5%
Table 49 – Composition of Government Nominees by NC (female only).
Gov AE/P EDU TL RE/S
2012 0.0% 0.0% 60.0% 40.0%
2013 0.0% 0.0% 25.0% 75.0%
2014 0.0% 0.0% 20.0% 80.0%
2015 0.0% 0.0% 20.0% 80.0%
2016 0.0% 0.0% 28.6% 71.4%
2017 14.3% 0.0% 14.3% 71.4%
Mean 2.4% 0.0% 28.0% 69.6%
Table 50 – Composition of Other Nominees by NC (female only).
Other AE/P EDU TL RE/S
2012 N/A N/A N/A N/A
2013 0.0% 0.0% 100.0% 0.0%
2014 N/A N/A N/A N/A
2015 0.0% 0.0% 100.0% 0.0%
2016 0.0% 0.0% 0.0% 100.0%
2017 N/A N/A N/A N/A
Mean 0.0% 0.0% 66.7% 33.3%
Summarizing the female case (for the all/male case, see §7.4.4.2):
Similarly to the male case, at least half of industry female Nominees are in RE/S and
this holds through the years as well. The remaining half splits between AE/P and TL, but
not equally as for the male case. The 27% of industry women are nominated as AE/P
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
86
and only 17% as TLs. There has not been any EDU nomination for industry women
since 2012.
Similarly to the male case, the vast majority of academic female Nominees is and has
always been in RE/S (88%).
Similarly to the male case, around 2/3 of Government female nominees are in RE/S
(70%) and about 1/3 is in TL (28%). However, we note that in 2017 there has been a
great jump of nominations in AE/P at the expense of TL.
Differently to the male case, 2/3 of Other female nominees is in TL and 1/3 in RE/S.
However, the number of Other female nominees is too small to draw any firm
conclusion (only 5 nominations were submitted in 2012-2017).
7.4.4.11 Breakdown of nomination categories of female nominees by employment
affiliation type
The breakdown is shown in the next four tables, from Table 51 to Table 54. Given the small
number of female nominees in some EAT/NC as well as the availability of only 5 samples
(years), the averages reported for non-RE/S categories must be taken with a grain of salt.
Table 51 – Composition of AE/P Nominees by EAT (female only).
AE/P Acad Gov Industry Other
2012 0.0% 0.0% 100.0% 0.0%
2013 50.0% 0.0% 50.0% 0.0%
2014 0.0% 0.0% 100.0% 0.0%
2015 0.0% 0.0% 100.0% 0.0%
2016 33.3% 0.0% 66.7% 0.0%
2017 20.0% 20.0% 60.0% 0.0%
Mean 17.2% 3.3% 79.4% 0.0%
Table 52 – Composition of EDU Nominees by EAT (female only).
EDU Acad Gov Industry Other
2012 100.0% 0.0% 0.0% 0.0%
2013 100.0% 0.0% 0.0% 0.0%
2014 100.0% 0.0% 0.0% 0.0%
2015 100.0% 0.0% 0.0% 0.0%
2016 100.0% 0.0% 0.0% 0.0%
2017 100.0% 0.0% 0.0% 0.0%
Mean 100.0% 0.0% 0.0% 0.0%
Table 53 – Composition of TL Nominees by EAT (female only).
TL Acad Gov Industry Other
2012 33.3% 50.0% 16.7% 0.0%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
87
TL Acad Gov Industry Other
2013 20.0% 20.0% 40.0% 20.0%
2014 50.0% 50.0% 0.0% 0.0%
2015 0.0% 25.0% 50.0% 25.0%
2016 40.0% 40.0% 20.0% 0.0%
2017 42.9% 14.3% 42.9% 0.0%
Mean 31.0% 33.2% 28.3% 7.5%
Table 54 – Composition of RE/S Nominees by NC (female only).
RE/S Acad Gov Industry Other
2012 81.4% 4.7% 14.0% 0.0%
2013 84.1% 6.8% 9.1% 0.0%
2014 86.8% 7.5% 5.7% 0.0%
2015 83.0% 8.5% 8.5% 0.0%
2016 76.0% 10.0% 8.0% 6.0%
2017 81.5% 7.7% 10.8% 0.0%
Mean 82.1% 7.5% 9.3% 1.0%
Summarizing the female case (for the all/male case, see §7.4.4.5):
Similarly to the male case, the vast majority of AE/P female Nominees have
traditionally been from the industry (79.4%) with academics being the second largest
group (17.2%). However, we note that in 2017 Government female Nominees in AE/P
jumped from zero to 20% mostly at the expense of Academics.
All EDU female Nominees are solely from academia in 2012-2017. For the male case,
EDU nominees from academia were 95.3%.
Differently from the male case where the TL dominant group was from industry
nominees, female TL nominees are uniformly distributed across academics,
Government, and industry. However, a dip of female nominees from Government in
2017 has left academics and industry female nominees equally split at around 40%.
Similarly to the male case, data has been rather stable in the past decade and academics
constitute the largest group at 82.1%, followed by industry and government Nominees at
9.3% and 7.5%, respectively.
7.4.4.12 Elevation probabilities segmenting data by nomination categories (female
only)
The available data is shown in Table 55. All the results discussed in this section can be obtained
starting from this table. Since non-RE/S nominees are very few and only five samples (years)
are available, the average probabilities for the non-RE/S case must be taken with a grain of salt.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
88
Table 55 – Nominations and elevations by NC (female only).
Nominations Elevations
Class AE/P EDU TL RE/S Total AE/P EDU TL RE/S Total
2012 1 2 6 43 52 0 1 1 21 23
2013 2 5 5 44 56 0 2 1 16 19
2014 3 3 2 53 61 2 1 0 16 19
2015 3 5 4 47 59 2 1 0 23 26
2016 3 2 5 50 60 0 2 2 19 23
Mean 2.4 3.4 4.4 47.4 57.6 0.8 1.4 0.8 19.0 22.0
Following the same methodology of §7.4.4.3 and §7.4.4.6, we obtain the marginal probabilities
of Table 56 and the direct and reverse conditional probabilities of Table 57.
Table 56 – Marginal probabilities of successful and unsuccessful nominations (female only), when
partitioning by NC. The last two-rows show column-wise (temporal) averages and the 95%-CI of the
average, respectively.
Class P(F) P(P) P(AEP) P(EDU) P(TL) P(RES)
2012 55.8% 44.2% 1.9% 3.8% 11.5% 82.7%
2013 66.1% 33.9% 3.6% 8.9% 8.9% 78.6%
2014 68.9% 31.1% 4.9% 4.9% 3.3% 86.9%
2015 55.9% 44.1% 5.1% 8.5% 6.8% 79.7%
2016 61.7% 38.3% 5.0% 3.3% 8.3% 83.3%
Mean 61.7% 38.3% 4.1% 5.9% 7.8% 82.2%
95%-CI ±7.3% ±7.3% ±1.7% ±3.3% ±3.8% ±4.1%
It is interesting to note that the largest number of Nominees is nominated in the RE/S category,
similar to the male case. The number of female nominees in other-than-RE/S NCs is much
smaller and somewhat close to the male case as well.
Table 57 – Direct and reverse conditional elevation probabilities, when partitioning by NC (female only).
The last two-rows show column-wise (temporal) averages and the 95%-CI of the average, respectively.
P(P|AEP) P(P|EDU) P(P|TL) P(P|RES) P(AEP|P) P(EDU|P) P(TL|P) P(RES|P)
2012 0.0% 50.0% 16.7% 48.8% 0.0% 4.3% 4.3% 91.3%
2013 0.0% 40.0% 20.0% 36.4% 0.0% 10.5% 5.3% 84.2%
2014 66.7% 33.3% 0.0% 30.2% 10.5% 5.3% 0.0% 84.2%
2015 66.7% 20.0% 0.0% 48.9% 7.7% 3.8% 0.0% 88.5%
2016 0.0% 100.0% 40.0% 38.0% 0.0% 8.7% 8.7% 82.6%
Mean 26.7% 48.7% 15.3% 40.5% 3.6% 6.5% 3.7% 86.2%
95%-CI ±45.3% ±38.1% ±20.6% ±10.2% ±6.3% ±3.6% ±4.6% ±4.5%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
89
The 95%-CI for the direct conditional elevation probability of female nominees is much wider
than for the all/male case (see Table 27) so that it is difficult to draw conclusions on it – for this
reason, we skip here the analysis of the edge metric. On the other hand, the 95%-CI of the
reverse conditional probabilities is smaller than the direct ones but still not small so that we can
state that female nominees are elevated somewhat proportionally to the a priori distribution of
female nominees across NCs, with the exception of TL nominees for which there are much
fewer elevations than one would expect given the number of TL nominations.
7.4.4.13 Elevation probabilities segmenting data by employment affiliation type
(female only)
The available data is shown in Table 58. All the results discussed in this section can be obtained
starting from this table.
Table 58 – Nominations and elevations by EAT (female only).
Nominations Elevations
Class Acad Govt Ind Oth Total Acad Govt Ind Oth Total
2012 39 5 8 0 52 19 1 3 0 23
2013 44 4 7 1 56 16 2 1 0 19
2014 50 5 6 0 61 15 0 4 0 19
2015 44 5 9 1 59 20 2 4 0 26
2016 43 7 7 3 60 18 3 2 0 23
Mean 44 5.2 7.4 1.0 57.6 17.6 1.6 2.8 0 22.0
Following the same methodology of §7.4.4.3 and §7.4.4.6, we obtain the marginal probabilities
of Table 59 and the direct and reverse conditional probabilities of Table 60.
Table 59 – Marginal probabilities of successful and unsuccessful nominations (female only), when
partitioning by EAT. The last row shows column-wise (temporal) averages.
Class P(F) P(P) P(Acad) P(Govt) P(Ind) P(Oth)
2012 55.8% 44.2% 75.0% 9.6% 15.4% 0.0%
2013 66.1% 33.9% 78.6% 7.1% 12.5% 1.8%
2014 68.9% 31.1% 82.0% 8.2% 9.8% 0.0%
2015 55.9% 44.1% 74.6% 8.5% 15.3% 1.7%
2016 61.7% 38.3% 71.7% 11.7% 11.7% 5.0%
Mean 61.7% 38.3% 76.4% 9.0% 12.9% 1.7%
95%-CI ±7.3% ±7.3% ±4.9% ±2.1% ±3.0% ±2.5%
It is interesting to note that the largest group of Nominees has become the academic one which
in 2016 accounted for 71.7% of all nominations, a percentage very close to the 71.1% of the
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
90
male case. Differently from the male case, there are fewer industry female Nominees and more
Government female nominees.
Table 60 – Direct and reverse conditional elevation probabilities, when partitioning by EAT (women only).
The last row shows column-wise (temporal) averages.
Pr(P|Aca) Pr(P|Gov) Pr(P|Ind) Pr(P|Oth) Pr(Aca/P) Pr(Gov/P) Pr(Ind/P) Pr(Oth/P)
2012 48.7% 20.0% 37.5% N/A 82.6% 4.3% 13.0% 0.0%
2013 36.4% 50.0% 14.3% 0.0% 84.2% 10.5% 5.3% 0.0%
2014 30.0% 0.0% 66.7% N/A 78.9% 0.0% 21.1% 0.0%
2015 45.5% 40.0% 44.4% 0.0% 76.9% 7.7% 15.4% 0.0%
2016 41.9% 42.9% 28.6% 0.0% 78.3% 13.0% 8.7% 0.0%
Mean 40.5% 30.6% 38.3% 0.0% 80.2% 7.1% 12.7% 0.0%
95%-CI ±9.2% ±25.3% ±24.2% 0.0% ±3.8% ±6.4% ±7.6% 0.0%
The 95%-CI of the conditional probabilities for non-academic female nominees is much wider
than for the all/male case (see Table 38) so that it is difficult to draw conclusions on it – for this
reason, we skip here the analysis of the edge metric.
7.4.4.14 Elevation probabilities segmenting data by both employment affiliation
type and nomination category (female only)
In this Section, we will consider subsets of female Nominees that were nominated in a given
nomination category (NC) and also have a given employment affiliation type (EAT). Available
data is for year 2012-2016, and the average number of Nominees (failed and passed) segmented
in both categories is shown in Table 61. Double segmentations further reduces the sample size,
so the results of this section must be taken with a grain of salt.
Table 61 – Average (2012-2016) number of failed and passed Nominees categorized by both NCs and EATs
(female only).
Average
2007-2016
Failed Nominees Passed Nominees
AE/P EDU TL RE/S AE/P EDU TL RE/S
Industry 1.2 0 1 2.4 0.8 0 0.2 1.8
Academics 0.4 2 0.8 23.2 0 1.4 0.4 15.8
Govt 0 0 1.4 2.2 0 0 0.2 1.4
Other 0 0 0.4 0.6 0 0 0 0
Calculating the conditional probabilities of elevation { | } for each
year, we obtain the minimum, mean, and maximum (over 2012-2016) elevation probabilities
shown in Table 62.
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
91
Table 62 – Conditional probabilities of elevation { | }, min|mean|max for 2012-
2016. The last column on the right contains { | } (see Table 60) while the bottom row
contains { | } (see Table 57), both averaged over 2007-2016.
AE/P EDU TL RE/S
Ind 0.0%|16.0%|40% 0.0%| 0.0%|12.5%|50% 20.0%|30.3%|40.0% 38.3%
Acad 0.0% 16.7%|30.7%|50.0% 0.0%|20.8%|50% 23.3%|28.9%|34.0% 40.5%
Govt 0.0% N/A 0.0%|6.7%|33.3% 0.0%|27.0%|40.0% 30.6%
Other 0.0% 0.0% 0.0% 0.0% 0.0%
26.7% 48.7% 15.3% 40.5%
The table shows several interesting things:
None of Other female nominees have ever been elevated, and no Government female
nominee in EDU was ever nominated (in 2012-2016),
The only female nominees elevated as AE/P are from industry and the only female
nominees elevated in EDU are from academia.
7.4.4.15 An interesting property unveiled in the analysis
We conclude pointing out that we have uncovered an interesting property that was never noticed
before. When considering all Nominees (male, female, undisclosed gender), the average
probabilities of elevation conditional to the EAT of the Nominee are very close to each other
and they are also very close to the average unconditional probability of being elevated (a priori
information). Despite the fact that this does not hold for every individual year, on the average it
does hold pretty well.
On the other hand, this does not hold true for all the average elevation probabilities conditional
to the NC. In this case, we note that while it holds for TL and RE/S, the average conditional
probabilities of AE/P and especially EDU strongly differ from the unconditional one. This is not
surprising as AE/P and EDU were already identified in §7.4.4.3 as the categories with the
lowest conditional elevation probability.
This can be seen in Table 63, where we tabulate the conditional and unconditional probabilities
of successful nominations. The last “Delta” row shows the differences between the conditional
probabilities and the unconditional probability of elevation. For all EATs as well as for RE/S
and TL, the average conditional probabilities are all within a few percent points of the
unconditional one.
Table 63 –Direct conditional and unconditional elevation probabilities.
P(P|Aca) P(P|Gov) P(P|Ind) P(P|Oth) P(P) P(P|AEP) P(P|EDU) P(P|TL) P(P|RES)
2007 31.7% 41.5% 42.8% 37.5% 35.0% 34.1% 23.4% 29.9% 36.9%
2008 37.5% 33.3% 41.2% 35.3% 38.2% 35.7% 41.2% 40.0% 37.9%
2009 39.8% 31.3% 42.9% 33.3% 39.9% 36.5% 25.0% 41.8% 40.9%
2010 38.0% 39.5% 38.5% 56.0% 38.8% 29.0% 23.3% 33.8% 41.5%
2011 38.2% 32.7% 45.2% 41.2% 39.5% 34.8% 28.6% 30.5% 41.7%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
92
2012 41.2% 35.2% 43.3% 41.7% 41.2% 33.3% 24.4% 35.4% 43.8%
2013 33.7% 36.9% 42.3% 27.8% 35.7% 28.1% 21.6% 38.8% 36.9%
2014 32.6% 42.6% 34.7% 68.8% 34.4% 32.9% 21.1% 42.9% 34.0%
2015 34.1% 40.4% 34.2% 23.1% 34.3% 28.6% 17.4% 30.1% 36.5%
2016 37.0% 38.2% 32.0% 14.3% 35.7% 25.0% 18.9% 38.2% 37.1%
Mean 36.4% 37.2% 39.7% 37.9% 37.3% 31.8% 24.5% 36.1% 38.7%
Delta -0.9% -0.1% 2.4% 0.6% 0.0% -5.5% -12.8% -1.1% 1.5%
This result can be expressed mathematically as:
{ | } { | } { | } { | }
{ | } { | } { },
The above tells us that the “elevation” event is statistically independent of the “belonging to a
given employment affiliation type” event.
If these two events were independent, then it would also follow that:
{ | } { }
{ | } { }
{ | } { }
{ | } { }
If the conditional probabilities of having an elevated Nominee belonging to EAT=X is
approximately equal to the unconditional probability of finding a Nominee (elevated or not)
belonging to the same X, then it also means that elevations are being made proportionally to the
a priori distribution of Nominees. Data confirms the above conclusion and also that the
conditional probabilities of having an elevated Nominee belonging to a given NC is
approximately equal to the unconditional probability of finding a Nominee (elevated or not)
belonging to that NC, as shown in Table 64.
Table 64 – Reverse conditional and unconditional elevation probabilities for (a) EATs and (b) NCs.
P(Acad/P) P(Gov/P) P(Ind/P) P(Oth/P) P(Acad) P(Govt) P(Ind) P(Oth)
2007 62.3% 10.1% 26.5% 1.1% 68.8% 8.5% 21.7% 1.0%
2008 63.7% 5.8% 28.5% 2.0% 64.8% 6.6% 26.4% 2.2%
2009 67.5% 5.0% 25.8% 1.7% 67.6% 6.3% 24.0% 2.0%
2010 66.7% 5.5% 23.3% 4.5% 68.0% 5.4% 23.5% 3.1%
2011 65.7% 5.6% 26.5% 2.2% 68.0% 6.8% 23.1% 2.1%
2012 71.1% 5.8% 21.6% 1.5% 71.2% 6.8% 20.6% 1.5%
2013 64.3% 8.1% 25.9% 1.7% 68.1% 7.8% 21.9% 2.2%
2014 65.5% 7.8% 22.9% 3.8% 69.1% 6.3% 22.7% 1.9%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
93
P(Acad/P) P(Gov/P) P(Ind/P) P(Oth/P) P(Acad) P(Govt) P(Ind) P(Oth)
2015 70.3% 7.0% 21.7% 1.0% 70.8% 5.9% 21.7% 1.5%
2016 73.7% 7.1% 18.5% 0.7% 71.1% 6.6% 20.6% 1.7%
Mean 67.1% 6.8% 24.1% 2.0% 68.8% 6.7% 22.6% 1.9%
Deltas -1.7% 0.1% 1.5% 0.1%
(a)
P(AEP/P) P(EDU/P) P(TL/P) P(RES/P) P(AEP) P(EDU) P(TL) P(RES)
2007 5.6% 4.1% 10.8% 79.5% 5.8% 6.1% 12.7% 75.4%
2008 6.8% 4.7% 12.2% 76.3% 7.2% 4.4% 11.6% 76.7%
2009 7.6% 3.0% 10.9% 78.5% 8.3% 4.8% 10.4% 76.5%
2010 5.8% 3.2% 8.4% 82.5% 7.8% 5.4% 9.7% 77.2%
2011 7.2% 3.1% 7.8% 81.9% 8.1% 4.3% 10.1% 77.5%
2012 4.3% 3.0% 10.3% 82.4% 5.3% 5.1% 12.0% 77.6%
2013 6.1% 2.7% 12.8% 78.5% 7.7% 4.5% 11.8% 76.1%
2014 8.5% 1.4% 11.3% 78.8% 8.9% 2.2% 9.0% 79.8%
2015 5.3% 2.7% 9.3% 82.7% 6.4% 5.3% 10.6% 77.7%
2016 4.4% 2.4% 11.4% 81.8% 6.2% 4.4% 10.7% 78.6%
Mean 6.2% 3.0% 10.5% 80.3% 7.2% 4.7% 10.9% 77.3%
Deltas -1.0% -1.6% -0.3% 3.0%
(b)
Let us now investigate if the same properties hold for the case of nominees segmented by
gender. If we segment data based only on the gender of the nominee, we can combine Table 44
and Table 45 into Table 65 where we can appreciate that:
The direct conditional elevation probability of male and female nominees is extremely
close to the unconditional probability of elevation, thus suggesting that the “elevation
event” for a nominee is independent of gender. We note however that the 95%-CI for the
direct conditional elevation of women is substantial at 6.1% so that the claim holds more
loosely for women.
The reverse conditional elevation probabilities are extremely close to the a priori
distribution of men and women, thus suggesting that elevations are being made
proportionally to the a priori distribution of nominated men and women. The claim can
be made with confidence as the 95%-CI for both men and women are rather small in this
case.
Table 65 – Summary of direct and reverse conditional and unconditional probabilities.
Class of P(P|Male) P(P|Female) P(P) P(Male|P) P(Male) P(Female|P) P(Female)
1999 41.5% 61.9% 42.2% 94.6% 96.3% 5.4% 3.7%
2000 46.9% 33.3% 46.7% 99.2% 98.9% 0.8% 1.1%
2001 49.4% 29.4% 48.8% 98.0% 96.8% 2.0% 3.2%
IEEE Fellow Strategic Planning Subcommittee – October 2016 Report
94
Class of P(P|Male) P(P|Female) P(P) P(Male|P) P(Male) P(Female|P) P(Female)
2002 46.2% 46.4% 46.2% 95.0% 95.0% 5.0% 5.0%
2003 41.2% 43.8% 41.3% 94.6% 94.9% 5.4% 5.1%
2004 39.7% 16.7% 38.5% 97.7% 94.7% 2.3% 5.3%
2005 34.3% 37.0% 34.4% 93.7% 94.1% 6.3% 5.9%
2006 35.6% 15.9% 34.5% 97.4% 94.4% 2.6% 5.6%
2007 34.4% 37.5% 34.6% 93.2% 93.7% 6.8% 6.3%
2008 36.8% 57.4% 38.0% 90.8% 93.9% 9.2% 6.1%
2009 39.2% 41.3% 39.4% 93.6% 93.9% 6.4% 6.1%
2010 38.2% 38.6% 38.3% 92.8% 92.8% 7.2% 7.2%
2011 38.1% 55.8% 39.2% 90.9% 93.6% 9.1% 6.4%
2012 41.0% 44.2% 41.2% 92.9% 93.4% 7.1% 6.6%
2013 35.7% 33.9% 35.6% 93.5% 93.2% 6.5% 6.8%
2014 34.6% 31.1% 34.3% 93.4% 92.7% 6.6% 7.3%
2015 33.3% 44.1% 34.1% 91.1% 93.1% 8.9% 6.9%
2016 35.1% 38.3% 35.3% 92.0% 92.6% 8.0% 7.4%
Mean 39.0% 39.3% 39.0% 94.1% 94.3% 5.9% 5.7%
95%-CI ±2.3% ±6.1% ±2.3% ±1.2% ±0.8% ±1.2% ±0.8%
Analysis of S/TC data
In this section, we applied the methodology used for the analysis of the IEEE data in §7.4.4 to
the following S/TCs: Computer Society (COMP), Communications Society (ComSoc), Power
& Energy Society (PES), and Signal Processing Society (SPS). For the sake of brevity,
however, data will be segmented in terms of EAT and only the metric EDGE ( ) will be
reported. This is shown in Table 66 to Table 69.
Table 66 – The edge for all EATs, COMP data only.
COMP Ind Edu Gov+Oth
2012 0.3 0.0 -0.7
2013 0.1 0.0 -0.2
2014 0.2 0.0 -0.4
2015 -0.2 0.1 0.1
2016 -0.1 0.3 -0.4
Mean 0.07 0.09 -0.32
STD 0.22 0.14 0.30
Table 67 – The edge for all EATs, ComSoc data only.
ComSoc Ind Edu Gov+Oth
2011 0.6 -0.1 -0.5
2012 -0.2 0.3 -0.4
2013 1.9 -0.5 -1.0
Recommended